JP2013077306A - Targeted object recognition device and targeted object recognition method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a barcode recognition device and barcode recognition method capable of efficiently executing barcode recognition processing with high accuracy.SOLUTION: A barcode recognition device 1 has an image interface 11, an image analysis part 12, an image conversion part 13, a bar recognition part 14, or the like. The image interface 11 acquires an image including a barcode photographed by a camera 2. The image analysis part 12 analyzes features in an input image acquired from the camera 2 and determines an image conversion method for converting the input image into an image for recognition processing on the basis of the analysis result. The image conversion part 13 converts the input image into the image for recognition processing by the image conversion method determined by the image analysis part 12. The bar recognition part 14 performs barcode recognition processing of the image for recognition processing acquired by the image conversion part 13.

Description

  The present invention is applied to, for example, a target recognition device that recognizes a detection target such as a character, a symbol, or a barcode attached to a medium such as a paper sheet, a document, or an article, and the target recognition device. It is related with the subject recognition method.

Conventionally, some systems that process paper sheets, documents, articles, and the like use bar code recognition processing or character recognition processing (OCR). For example, in a sorting system that sorts sorting objects such as paper sheets, documents, or articles, the sorting processing is performed by recognizing characters, symbols, or barcodes indicating sorting information printed on the sorting objects. There is. In addition, in the sorting system for sorting mail items, the destination information indicated by characters or symbols is recognized by the character recognition processing, and the destination information obtained as a recognition result is converted into a barcode and printed on the mail piece. Has been put to practical use. That is, in the sorting system as described above, a character recognition device that recognizes characters or symbols or a barcode recognition device that recognizes barcodes is mounted.
For example, the barcode recognition apparatus as described above performs a recognition process (barcode recognition process) for recognizing a barcode from an image including a barcode print area photographed by a camera. Further, the character recognition apparatus as described above performs a recognition process (character recognition process) for recognizing a character or a symbol from an image including a character or symbol description area photographed by the camera. A recognition program for performing such recognition processing of a detection target is normally configured to process an image having a predetermined number of gradations at the design stage. On the other hand, in recent years, the photographing performance of a camera for photographing a character / symbol description region or a barcode print surface has been improved. For example, an image taken by a camera has become an image with a large amount of information such as the number of gradations as the performance of the camera improves.

  In such a situation, when the camera for capturing an image to be recognized is changed to a camera for capturing a high gradation image, the high gradation image and the low gradation image captured by the camera are processed. The challenge is to efficiently collaborate with existing recognition programs. For example, it is conceivable to change an existing recognition program to one that processes a high gradation image captured by a camera. In this case, it takes time to modify the recognition program and the like, and since the image having a large amount of information is a processing target, the recognition processing itself after the modification tends to be heavy. It is also conceivable to convert a high gradation image captured by the camera into a low gradation image in accordance with the recognition program by a predetermined conversion method at a stage prior to the recognition process. In this case, although the performance of the camera has been improved, the recognition process cannot be improved in accuracy.

JP-A-2005-45606

  An object of one aspect of the present invention is to provide an object recognition device and an object recognition method that can execute recognition processing of a detection object efficiently and with high accuracy.

  An object recognition apparatus according to an aspect of the present invention includes an image acquisition unit that acquires an input image including an image to be detected, and a first process that performs detection target recognition processing on the input image acquired by the image acquisition unit. And when the recognition of the detection target by the first detection target recognition unit fails, based on the information obtained in the process of the recognition process of the detection target in the first detection target recognition unit An image analysis unit for determining an image conversion method for reconverting an input image into an image for recognition processing, and image conversion for converting the input image into an image for recognition processing by the image conversion method determined by the image analysis unit And a second detection target recognition unit that performs a recognition target recognition process on the recognition processing image obtained by the image conversion unit.

  A subject recognition method according to one aspect of the present invention is a method used in a subject recognition device, which obtains an input image including a detection target image and recognizes the detection target with respect to the input image. When the recognition of the detection target by the first recognition process fails, the input image is converted into an image for recognition process based on the information obtained in the process of the first recognition process. An image conversion method is determined, the input image is converted into an image for recognition processing by the determined image conversion method, and a detection target is detected for the image for recognition processing obtained by image conversion by the image conversion method A second recognition process for recognizing is performed.

FIG. 1 is a diagram illustrating a configuration example of a barcode recognition apparatus. FIG. 2 is a diagram illustrating a schematic configuration example of a sorting system to which the barcode recognition device is applied. FIG. 3 is a diagram illustrating an example of an image captured by the camera. FIG. 4 is a diagram for explaining an example of image conversion processing. FIG. 5 is a diagram showing characteristics (conversion tables) of the first to fifth image conversion processes. 6A to 6E are diagrams illustrating examples of images obtained by the first to fifth image conversion processes. FIG. 7 is a flowchart for explaining the flow of the first processing example in the barcode recognition apparatus. FIG. 8 is a diagram illustrating an example of a density histogram for all pixels of an input image having 4096 gradations. FIG. 9 is a diagram for explaining the medium density value determined by the percentile method. FIG. 10 is a diagram for explaining an example of the conversion table corresponding to the medium density value. FIG. 11 is a flowchart for explaining the flow of the second processing example in the barcode recognition apparatus. FIG. 12 is a diagram for explaining the bar density value determined by the percentile method. FIG. 13 is a diagram for explaining an example of the conversion table corresponding to the bar density value. FIG. 14 is a flowchart for explaining the flow of the third processing example in the barcode recognition apparatus. FIG. 15 is a flowchart for explaining the flow of the fourth processing example in the barcode recognition apparatus. FIG. 16 is a diagram for explaining an example of the conversion table based on the separation threshold. FIG. 17 is a flowchart for explaining the flow of the fifth processing example in the barcode recognition apparatus.

Embodiments of the present invention will be described below with reference to the drawings.
In the following embodiments, a barcode recognition device and a barcode recognition method will be described as examples of the subject recognition device and the subject recognition method. However, the bar code recognition device and the bar code recognition method described below are also applicable to a character recognition device and a character recognition method as other examples of the target recognition device and the target recognition method.

FIG. 1 is a diagram showing a configuration example of a barcode recognition apparatus 1 according to an embodiment of the present invention.
Here, as an example of the barcode recognition device 1, it is assumed that a barcode printed with fluorescent ink on a printing surface of a paper sheet as a medium is recognized.
In the example shown in FIG. 1, the barcode recognition apparatus 1 includes an image interface 11, an image analysis unit 12, an image conversion unit 13, a bar recognition unit (detection target recognition unit) 14, a feedback unit 15, and an output interface 16. ing.

  The bar code recognition device 1 is an example of an object recognition device. The barcode recognition apparatus 1 as the object recognition apparatus is realized by, for example, an electronic computer (computer) having a function of performing information processing using various programs. The computer as the barcode recognition apparatus 1 includes an image input interface corresponding to the image interface 11, an interface for outputting a recognition result corresponding to the output interface 16, and control for executing various processes. It consists of parts. The control unit of the computer as the barcode recognition device 1 includes a CPU, a RAM, a ROM, a rewritable nonvolatile memory, and the like. In such a control unit, the CPU realizes various processes by executing various programs stored in a ROM or a nonvolatile memory using a RAM or the like as a working memory.

  For example, a control unit of a computer as the barcode recognition device 1 functions as the image analysis unit 12 by executing a program for image analysis. Further, the control unit of the computer as the barcode recognition device 1 functions as the image conversion unit 13 by executing a program for image conversion. The control unit of the computer as the barcode recognition device 1 functions as the barcode recognition unit 14 by executing a barcode recognition processing program. The control unit of the computer as the barcode recognition device 1 functions as the feedback unit 15 by executing a program for feedback processing. The image analysis unit 12, the image conversion unit 13, the bar recognition unit 14, and the feedback unit 15 may be realized by hardware such as an integrated circuit.

  Further, the camera 2 is connected to the barcode recognition device 1. The camera 2 reads an image of a surface on which a barcode is printed on a paper sheet. For example, in the form of recognizing a barcode printed on a paper sheet with fluorescent ink, the camera 2 includes a fluorescent scanner and illumination. In this case, the illumination irradiates the surface on which the barcode of the paper sheet is printed with light for exciting the phosphor contained in the fluorescent ink. For example, the illumination is composed of a fluorescent lamp or an LED. The fluorescence scanner receives fluorescence emitted from a surface on which a barcode of a paper sheet irradiated with light from the illumination is printed. That is, the fluorescent scanner acquires a barcode printed with fluorescent ink on the surface of a paper sheet as image information (fluorescent image). The fluorescence image read by the fluorescence scanner is supplied to the image interface (image acquisition unit) 11 of the barcode recognition device 1.

  The camera 2 captures an image having a larger amount of information than an image to be processed by the bar recognition unit 14 in the barcode recognition apparatus 1. For example, the camera 2 captures an image (for example, a 12-bit image, a 16-bit image) having a higher gradation than the gradation (for example, an 8-bit image) of the image to be processed by the bar recognition unit 14. . In the following description, an image (input image) captured by the camera 2 is a 12-bit image with 4096 gradations, and an image (an image for recognition processing) that is processed by the bar recognition unit 14 is mainly used. Assume an 8-bit image with 256 gradations.

  The image analysis unit 12 or the image conversion unit 13 may be provided in the camera 2. In this case, the barcode recognition apparatus 1 may be configured to supply an image (8-bit image) after image conversion by the camera to the bar recognition unit via the image interface. Also, in such a configuration, for example, the control unit in the camera 2 executes a program for image analysis, thereby realizing the same function as the image analysis unit 12 or executing a program for image conversion. By doing so, the same function as the image conversion unit 13 can be realized.

  The image interface 11 functions as an image acquisition unit that acquires an image to be recognized (hereinafter also simply referred to as an input image) including a barcode image as a detection target. That is, the image interface 11 is an interface for inputting an image captured by the camera 2. The input image acquired from the camera 2 by the image interface 11 is supplied to the image analysis unit 12 and the image conversion unit 13.

  The image analysis unit 12 performs processing for analyzing an image (input image) taken by the camera 2 acquired by the image interface 11 and determining an image conversion method. That is, the image analysis unit 12 performs a process for determining (estimating) a feature in the input image and a process for determining an image conversion method for converting the input image into an image for recognition processing. The image analysis unit 12 supplies information indicating a conversion method for converting the input image into an image for recognition processing to the image conversion unit 13. The image analysis unit 12 corresponds to the first and second image analysis units.

  In the example illustrated in FIG. 1, the image analysis unit 12 includes a density value estimation unit 21, a medium density estimation unit 23, a bar density estimation unit 24, and a conversion method determination unit 26.

  The density value estimator 21, the bar density estimator 24, and the medium density estimator 23 each perform a process of determining (estimating) a feature in the input image. For example, the density value estimation unit 21 determines a mode density value, an average density value, a percentile method density value, and the like in the input image as features in the input image. The bar density estimation unit 24 determines the density value of the detection element as the detection element image in the input image. That is, the bar density estimation unit 24 determines a density value of a bar image as a detection element image constituting a barcode as a detection target in the input image as a feature in the input image. The medium density estimation unit 23 determines the density value of the medium as a background image in the input image.

  The density value estimator 21, the bar density estimator 24 and the medium density estimator 23 determine information for determining an image conversion method. Therefore, any one of the density value estimation unit 21, the bar density estimation unit 24, and the medium density estimation unit 23 may be selectively provided. For example, when determining the image conversion method based on the characteristics (mode density value, average density value, density value by percentile method, etc.) obtained from the entire input image, the image analysis unit 12 uses the density value estimation unit 21. Should be provided. When determining an image conversion method based on a combination of the density value of each bar and the density value of the medium in the input image, the image analysis unit 12 includes the bar density estimation unit 24 and the medium density estimation unit 23. Can be provided.

  The conversion method determination unit 26 determines an image conversion method for converting an input image into an image for recognition processing. That is, the conversion method determination unit 26 recognizes the input image based on the analysis result of the feature in the input image obtained by the density value estimation unit 21, the bar density estimation unit 41, or the medium density estimation unit 42. The image conversion method for converting to is determined. For example, when the image for recognition processing to be processed by the bar recognition unit 14 is 8 bits and the input image captured by the camera 2 is 12 bits, the conversion method determination unit 22 determines the characteristics of the input image. Based on this, an image conversion method for converting a 12-bit input image into an 8-bit image is determined. Information indicating the image conversion method determined by the conversion method determination unit 22 is supplied to the image conversion unit 13.

  The image conversion unit 13 converts the input image captured by the image interface 11 into a recognition processing image to be processed by the bar recognition unit 14. The image conversion unit 13 has a function of performing image conversion processing by various conversion methods. The image conversion unit 13 converts the input image into an image for recognition processing by the image conversion method determined by the image analysis unit 12. The image conversion process by the image conversion unit 13 will be described in detail later.

  The bar recognition unit 14 functions as a detection target recognition unit, and performs a process for recognizing a barcode as a detection target. The bar recognition unit 14 performs a process of recognizing a barcode included in an image for recognition processing (converted image) supplied from the image conversion unit 13. The bar recognition unit 14 includes, for example, a bar detection unit, a decoding unit, and a determination unit. The bar recognition unit 14 corresponds to a first and second bar recognition unit.

  The bar detection unit functions as a detection target detection unit and performs a process of detecting each bar constituting a barcode as a detection target. The bar detection unit detects each bar constituting the barcode from the image for recognition processing supplied from the image conversion unit 13 as a bar candidate. The bar detection unit determines the type for each detected bar. The bar detection unit determines the type of each bar based on the length of each bar and the relative position in the entire bar code in accordance with the bar code specifications. The bar detection unit also has a function of detecting the entire barcode area as a barcode area candidate based on the position of each detected bar candidate.

For example, the bar detection unit detects and recognizes each bar according to the following procedure.
When an 8-bit image is supplied from the image conversion unit 13, the bar detection unit extracts pixels whose density value exceeds a predetermined threshold in the 8-bit image.
When a pixel whose density value exceeds a predetermined threshold is extracted, the bar detection unit selects a region where the extracted pixels are solid and present as a bar candidate.
When a bar candidate is selected, the bar detection unit extracts a part where the bar candidates are arranged in the horizontal direction (a direction orthogonal to the long side direction of each bar candidate) as a barcode area.
When the bar code area is extracted, the bar detection unit determines (estimates) the center position of each bar candidate in the vertical direction (long side direction of each bar candidate) in the bar code area.

When the center position in the vertical direction is determined, the bar detection unit encodes each bar candidate. That is, the bar detection unit determines the type of bar for each bar candidate based on the center position, and codes the determination result. For example, when the barcode to be recognized is an array of four types of bars, an ascender bar, a descender bar, a long bar, and a timing bar, the bar detection unit determines which type of bar each bar candidate is. judge. That is, the bar detection unit determines a bar candidate having a bar formed above the center position as an ascender bar, determines a bar candidate having a bar formed below the center position as a descender bar, and sets the center as a reference. The bar candidate in which a long bar is formed vertically is determined as a long bar, and the bar candidate in which a short bar is formed with reference to the center is determined as a timing bar.
When the bar detection unit determines the type of each bar candidate, the bar detection unit encodes the determination result of each bar candidate. The bar detection unit supplies information obtained by arranging the determination results of the encoded bar candidates in the order in which the bar candidates are arranged to the decoding unit as a barcode detection result.

  The decoding unit decodes the barcode detected by the bar detection unit. The decoding unit decodes the barcode based on the information indicating the type of each bar and the information indicating the arrangement of the bars as the barcode detection result by the bar detection unit. The decoding unit may have a function of correcting an error based on an error correction code described later. In this case, the decoding unit decodes information indicated by the barcode based on the detection result by the bar detection unit, and corrects the decoded information by error correction processing based on the error correction code.

  The determination unit determines whether or not the barcode decoding result (barcode recognition result) by the decoding unit is valid. For example, the determination unit determines whether the recognition result is valid based on whether the barcode recognition result is information of a predetermined number of digits. The determination unit may compare dictionary data stored in a dictionary database (not shown) with a barcode recognition result and determine that the recognition result existing as dictionary data is valid. When the determination unit determines that the barcode decoding result is valid, the determination unit outputs the decoding result obtained by the decoding unit to the output interface 16 as a barcode recognition result. If the determination unit determines that the barcode decoding result is not valid, the determination unit supplies the feedback unit 15 with the decoding result obtained by the decoding unit or the bar detection result by the bar detection unit.

  When the barcode recognition device 1 as the object recognition device is applied to a character recognition device, the bar recognition unit 14 corresponds to a character recognition unit that recognizes a character or a symbol by a predetermined recognition program. . In this case, the bar detection unit and the decoding unit correspond to a processing unit that detects character candidates and recognizes each character candidate, and the determination unit determines whether the character recognition result is valid. It may be considered that it corresponds to a processing unit.

  The feedback unit 15 feeds back the information obtained by the recognition processing by the bar recognition unit 14 to the image analysis unit 12. Here, the feedback unit 15 feeds back to the image analysis unit 12 information that can be used for image analysis among the information obtained in the recognition processing by the bar recognition unit 14. For example, the feedback unit 15 calculates a separation threshold value between the bar density and the background density based on the candidate bar information obtained by the bar detection unit of the bar recognition unit 14, and uses the calculated separation threshold value as an image. Feedback to the analysis unit 12. In this case, the image analysis unit 12 performs a process of analyzing the image again based on the information supplied from the feedback unit 15. This process will be described in detail later.

  The output interface 16 is an interface for outputting the barcode recognition result obtained by the bar recognition unit 14 to the outside. For example, when the determination unit of the bar recognition unit 14 determines that the barcode decoding result is valid, the output interface 16 outputs the decoding result obtained by the decoding unit to the outside as the barcode recognition result. To do. Further, if the output interface 16 determines that the determination is not valid by the determination unit of the bar recognition unit 14, the output interface 16 outputs to the outside that the barcode recognition has failed if the feedback processing by the feedback unit 15 is not performed. .

FIG. 2 is a diagram showing a schematic configuration of a sorting system to which the barcode recognition apparatus 1 is applied.
The sorting system shown in FIG. 2 is a system that sorts an article based on a bar code printed on the article (for example, paper sheets). Here, the article to be subjected to the sorting process will be described assuming paper sheets on which barcodes are printed with information (sorting information) indicating the sorting destination with fluorescent ink on the first surface. .
The sorting system shown in FIG. 2 includes a control device 31, a feeding device 32, a transport device 33, a sorting device 34, a camera 2, a barcode recognition device 1, and the like. In the configuration example shown in FIG. 2, a supply device 32, a transport device 33, a sorting device 34, a barcode recognition device 1, and the like are connected to the control device 31.

  The control device 31 is responsible for overall control of the sorting system. The control device 31 includes, for example, an electronic computer having a CPU, various memories, various interfaces, and the like. The control device 31 has a function of executing various processes when the CPU executes a program stored in the memory. For example, the control device 31 is connected to each unit through various interfaces, and performs operation control of each unit based on a control program stored in the memory. The control device 31 is also connected to an operating panel (not shown) that displays guidance for the operator and receives operation instructions from the operator.

  In the supply device 32, paper sheets to be subjected to sorting processing are set in a predetermined direction. The supply device 32 takes out the set paper sheets one by one and supplies them to the conveyance path of the conveyance device 33. The transport device 33 transports each paper sheet based on the control by the control device 31. For example, the transport device 33 transports each paper sheet sequentially taken out by the supply device 32 at a constant transport speed.

  An image of the first surface of the paper sheet transported by the transport device 33 is taken by the camera 2. As described above, the camera 2 includes an illumination and a fluorescent scanner, and captures an image including a barcode printed with fluorescent ink on the first surface of the paper sheet. The barcode recognition apparatus 1 performs barcode recognition processing on an image captured by the camera 2 with the above-described configuration. The barcode recognition result by the barcode recognition device 1 is supplied to the control device 31. The control device 31 performs processing for sorting the paper sheets by controlling the transport device 33 and the sorting device 34 based on the barcode recognition result by the barcode recognition device 1.

  The sorting device 34 has a plurality of sorting pockets for sorting and accumulating paper sheets. For example, the sorting device 34 has a plurality of sorting pockets, a plurality of gate mechanisms, a branch conveyance path, and the like partitioned into a plurality of stages and a plurality of rows. In this case, the sorting device 34 transports and accumulates each paper sheet into a desired sorting pocket by controlling each gate mechanism in accordance with a control signal from the control device 31.

Next, an image captured by the camera 2 will be described.
In the present embodiment, it is assumed that a barcode printed with fluorescent ink is recognized. For this reason, the image captured by the fluorescent scanner as the camera 2 is predicted to be an image with a bright bar and a dark background. For example, FIG. 3 is a diagram illustrating an example of an image captured by the camera 2. In the image shown in FIG. 3, the bar is bright and the background (medium) is dark. As shown in FIG. 3, in an image in which the bar and the background can be clearly distinguished, the bar can be reliably extracted, so that the barcode recognition accuracy is increased. On the other hand, if the bar and the background are both bright, or if the bar and the background are both dark, it is difficult to detect the bar, resulting in low barcode recognition accuracy.

  In the barcode recognition apparatus 1, the image analysis unit 12 analyzes the characteristics of the image (input image) captured by the camera 2, and the image conversion unit 13 converts the input image into a barcode recognition accuracy based on the analysis result. A process of converting to an image that becomes higher (for example, converting to an image in which a bar can be easily detected) is performed. Here, it is assumed that the camera 2 captures a 12-bit image and the bar recognition unit performs barcode recognition processing on the 8-bit image. That is, in this embodiment, it is assumed that a 12-bit input image is converted into an 8-bit recognition image based on the characteristics of the input image.

Next, the image conversion process in the image conversion unit 13 will be described.
FIG. 4 is a diagram for explaining an example of image conversion processing. In the example shown in FIG. 4, a 12-bit image (an image having 4096 gradations) in which the density value (luminance value) of each pixel is 0 to 4095 is used and the density value (luminance value) of each pixel is 0 to 255. Is linearly converted to an 8-bit image (256-gradation image) indicated by the value of. FIG. 4 shows that the higher the numerical value, the brighter, and the lower the numerical value, the darker. Hereinafter, in the present embodiment, as shown in FIG. 4, the higher the density value, the brighter the pixel.

  In the image conversion as shown in FIG. 4, an image in the 12-bit range can be converted into an image in the 8-bit range without unevenness. However, the image conversion as shown in FIG. 4 cannot be expected to improve the barcode recognition accuracy such as the bar detection accuracy. That is, in the 12-bit image, when the difference between the density value of each pixel constituting the bar and the density value of each pixel constituting the background is small (that is, when both the bar and the background are bright, or both the bar and the background are In the case of dark), in the 8-bit image obtained by the image conversion as shown in FIG. 4, it may be difficult to distinguish the bar from the background. This is because when the density difference between the background and the bar in the 12-bit image is small, the density difference may be further reduced or lost when the 12-bit image is converted to the 8-bit image.

Hereinafter, first to fifth processing examples will be described as processing examples for converting an image (also referred to as an input image) taken by the camera 2 into an image for recognition processing.
First, a first processing example will be described.
In this first processing example, the image conversion method is determined based on the characteristics (density value distribution) in the entire 4096-gradation 12-bit image. In the first processing example, the image conversion unit 13 performs the following five types of image conversion processing (first to fifth image conversion processing) as a method of converting a 12-bit image into an 8-bit image. It shall have a function. Here, examples of the first to fifth image conversion processes will be described.
FIG. 5 is a diagram showing characteristics (conversion tables) of the first to fifth image conversion processes. FIG. 6 is a diagram illustrating examples of images obtained by the first to fifth image conversion processes.

  (A) In the first image conversion process, a 12-bit image is linearly converted into an 8-bit image in which pixels having a density value of 0 to 255 are 256 gradation pixels. In this case, pixels having a density value of 256 or more in the 12-bit image are converted to a density value of 255. This first image conversion process is a conversion process according to the characteristic a shown in FIG. In the first image conversion process, a bright 8-bit image (256-gradation image) is obtained by emphasizing the gradation change by each pixel having a density value of 0 to 255 in the 12-bit image. For example, in the first image conversion process, a bright image as shown in FIG. Therefore, the first image conversion process is considered to be effective for an image conversion process for an input image having a large number of pixels having a density value of less than 255 (an image that is entirely dark). In particular, in the first image conversion process, the difference between the density distribution of each pixel constituting the background image and the density distribution of each pixel constituting the bar image in the 4096 gradation input image is a density value of 0-255. It is considered effective when it can be emphasized in the range of.

  (B) In the second image conversion process, each pixel having a density value of 0 to 511 in a 12-bit image is linearly converted into an 8-bit image having 256 gradations. In this case, the density value of each pixel having a density value of 512 or more in the 12-bit image is converted to 255. This second image conversion process is an image conversion process according to the characteristic b shown in FIG. In the second image conversion process, an 8-bit image (256-gradation image) is obtained by emphasizing a gradation change by each pixel having a density value of 0 to 511 in a 12-bit image (4096-gradation image). For example, in the second image conversion process, an image as shown in FIG. 6B is obtained. Therefore, the second image conversion process is considered to be effective for an image conversion process for an input image having many pixels having density values less than 511. That is, in the second image conversion process, the difference between the density distribution of each pixel constituting the background image and the density distribution of each pixel constituting the bar image in the 4096 gradation input image is a density value of 0 to 511. It is considered effective when the range can be emphasized.

  (C) The third image conversion processing linearly converts each pixel having a density value of 0 to 1023 into an 8-bit image having 256 gradations in a 12-bit image. In this case, the density value of each pixel having a density value of 1024 or more in the 12-bit image is converted to 255. This third image conversion process is an image conversion process according to the characteristic c shown in FIG. In the third image conversion process, an 8-bit image (256-gradation image) is obtained in which a gradation change by each pixel having a density value of 0 to 1023 is emphasized in a 12-bit image (4096-gradation image). For example, in the third image conversion process, an image as shown in FIG. 6C is obtained. Therefore, the third image conversion process is considered to be effective for an image conversion process for an input image having many pixels with density values less than 1023. That is, in the third image conversion process, the difference between the density distribution of each pixel constituting the background image and the density distribution of each pixel constituting the bar image in the 4096-gradation input image is a density value of 0 to 1023. It is considered effective when the range can be emphasized.

  (D) In the fourth image conversion process, each pixel having a density value of 0 to 2047 in a 12-bit image is linearly converted into an 8-bit image having 256 gradations. In this case, the density value of each pixel having a density value of 2048 or more in the 12-bit image is converted to 255. The fourth image conversion process is an image conversion process according to the characteristic d shown in FIG. In the fourth image conversion process, an 8-bit image (256-gradation image) is obtained in which a gradation change by each pixel having a density value of 0 to 2047 is emphasized in a 12-bit image (4096-gradation image). For example, in the fourth image conversion process, an image as shown in FIG. 6D is obtained. Therefore, the fourth image conversion process is considered to be effective for an image conversion process for an input image having many pixels with density values less than 2047. That is, in the fourth image conversion process, the difference between the density distribution of each pixel constituting the background image and the density distribution of each pixel constituting the bar image in the 4096-gradation input image is a density value of 0 to 2047. It is considered effective when the range can be emphasized.

  (E) The fifth image conversion process linearly converts each pixel having a density value of 0 to 4095 in a 12-bit image into an 8-bit image having 256 gradations. In this case, the density value of each pixel having a density value of 4095 or higher in the 12-bit image is converted to 255. The fifth image conversion process is an image conversion process according to the characteristic e shown in FIG. In the fifth image conversion process, as shown in FIG. 4, an 8-bit image that does not change the gradation change in the entire 12-bit image is obtained. For example, in the fifth image conversion process, an image as shown in FIG. 6E is obtained. Therefore, the fifth image conversion processing is performed in an input image in which the density value of each pixel is distributed in a wide range of 0 to 4095, or an input image having many bright pixels having a density value of 2047 or more (the entire image is bright). This is considered to be effective for image conversion processing.

  These first to fifth image conversion processes are processes executed by performing a bit shift operation on the density value of each pixel. That is, the above-described first to fifth image conversion processes are processes that do not require complicated arithmetic processing, and can obtain the result of the image conversion process at high speed. For this reason, the image conversion unit 13 may execute the first to fifth image conversion processes when an input image is given. In this case, the image conversion unit 13 can supply the bar recognition unit 14 with an image subjected to image conversion processing by the designated image conversion method when the image conversion method is designated by the image analysis unit 12.

The image analysis unit 12 analyzes a feature in the input image and determines one of the first to fifth image conversion processes based on the analysis result. For example, in the image analysis unit 12, the density value estimation unit 21 calculates the most frequent density value in the input image as a feature in the input image. It is determined that the density value (mode density value) having the highest frequency in the input image is the background density value. This is because it is assumed that the background around the area where the barcode is printed has a substantially uniform density value. That is, the mode density value is determined to be the density value (medium density value) having the highest frequency in the background image.
If the image conversion method is determined based on such a mode density value, the image analysis unit 12 can determine an image conversion method according to the density value of the background image. That is, when the image conversion unit 13 has the functions of the first to fifth image conversion processes, the conversion method determination unit 26 sets the mode in the input image calculated by the density value estimation unit 21. One of the first to fifth image conversion processes is selected based on the density value.

As a method of determining the image conversion method, for example, a method of determining the image conversion method by comparing a predetermined threshold value with the most frequent density value calculated by the density value estimation unit 21 can be considered. In this case, the threshold value for determining the image conversion method is appropriately set according to the operation mode.
For example, it is assumed that four threshold values “100”, “200”, “500”, and “1000” are set as threshold values for the most frequent density value in a 12-bit image. In this case, the conversion method determination unit 26 can determine one of the first to fifth image conversion processes as the image conversion method for the input image based on the following conditions.

(A-1) If the mode density value α is less than “100” (α <100), the first image conversion process is selected as the image conversion method.

(B-1) If the mode density value α is not less than “100” and less than “200” (100 ≦ α <100), the second image conversion process is selected as the image conversion method.

(C-1) If the mode density value α is “200” or more and less than “500” (200 ≦ α <500), the third image conversion process is selected as the image conversion method.

(D-1) If the mode density value α is not less than “500” and less than “1000” (500 ≦ α <1000), the fourth image conversion process is selected as the image conversion method.

(E-1) If the mode density value α is “1000” or more (1000 ≦ α), the fifth image conversion process is selected as the image conversion method.

  The density value estimation unit 21 determines an average density value in the input image or a density value determined (estimated) by the percentile method instead of the mode density value in the input image as described above. Also good. The average density value is an average value of density values of all pixels in the input image. Further, as the determination of the density value by the percentile method, it is possible to determine the density value when the value integrated in order from the darkest density value in the density histogram of the input image exceeds a predetermined threshold value.

  The conversion method determination unit 26 may select a plurality of image conversion methods. For example, the conversion method determination unit 26 may select the first image conversion process and the second image conversion process if α <100. In such a case, the image conversion unit 13 supplies the bar recognition unit 14 with the first image obtained by the first image conversion process and the second image obtained by the second image conversion process. To do. Therefore, the bar recognition unit 14 may perform barcode recognition processing on the first image and the second image.

Next, the flow of processing according to the first processing example as described above will be described.
FIG. 7 is a flowchart for explaining the flow of the first processing example in the barcode recognition apparatus 1. Here, as an example, as described above, four threshold values (“100”, “200”, “500”, “1000”) are set for the mode density value α, and based on these threshold values, the above-described threshold values are set. A description will be given assuming that one of the first to fifth image conversion processes is selected as the image conversion method.

  First, it is assumed that the camera 2 captures the barcode print surface with 12 bits (4096 gradations). Then, the image interface 11 of the barcode recognition device 1 acquires a 12-bit image captured by the camera 2 (step S10). The image interface 11 supplies the acquired 12-bit image to the image analysis unit 12 and the image conversion unit 13. When a 12-bit image is supplied from the image interface 11, the image analysis unit 12 determines the most frequent density value α in the 12-bit image by the density value estimation unit 21 (step S11).

  When the mode value estimator 21 determines the mode density value α, the conversion method determination unit 26 selects one of the first to fifth image conversion processes as the mode conversion method for the 12-bit image. A selection is made based on the density value α and a predetermined threshold (steps S12 to S20).

  For example, when the mode density value α is less than 100 (step S12, YES), the conversion method determination unit 26 uses the first image conversion process (density value of 4096 gradations) as an image conversion method for the 12-bit image. A process of linearly converting 0 to 255 into 255 gradation density values 0 to 255 is selected (step S13). When the most frequent density value α is 100 or more and less than 200 (step S14, YES), the conversion method determination unit 26 uses the second image conversion process (density of 4096 gradations) as an image conversion method for the 12-bit image. A process of linearly converting the values 0 to 511 into 255 gradation density values 0 to 255 is selected (step S15).

  If the most frequent density value α is 200 or more and less than 500 (step S16, YES), the conversion method determination unit 26 uses the third image conversion process (density of 4096 gradations) as the image conversion method for the 12-bit image. A process of linearly converting the values 0 to 1023 into the density values 0 to 255 of 255 gradations is selected (step S17). When the mode density value α is 500 or more and less than 1000 (step S18, YES), the conversion method determination unit 26 uses a fourth image conversion process (density of 4096 gradations) as an image conversion method for the 12-bit image. A process of linearly converting the value 0 to 2047 into the density value 0 to 255 of 255 gradations is selected (step S19). When the mode density value α is 1000 or more (step S18, NO), the conversion method determination unit 26 uses the fifth image conversion process (4096 gradation density values 0 to 0) as an image conversion method for the 12-bit image. The process of linearly converting 4095 to 255 gradation density values 0 to 255 is selected (step S20).

When an image conversion method for a 12-bit image is selected in steps S12 to S20, the conversion method determination unit 26 designates the selected image conversion process to the image conversion unit 13 (step S21).
When an image conversion method (image conversion process) for the 12-bit image is specified by the conversion method determination unit 26 of the image analysis unit 12, the image conversion unit 13 converts the 12-bit image into an 8-bit image by the specified image conversion process. (Step S22). The image conversion unit 13 supplies the bar recognition unit 14 with the 8-bit image generated by the designated image conversion process. The image conversion unit 13 may execute the first to fifth image conversion processes on the 12-bit image acquired from the image interface 11 in parallel with the process in the image analysis unit 12. . In this case, the image conversion unit 13 supplies the 8-bit image generated by the image conversion method designated by the image analysis unit 12 to the bar recognition unit 14.

  The bar recognition unit 14 performs barcode recognition processing on the 8-bit image supplied from the image conversion unit 13 (step S23). In the barcode recognition process in the bar recognition unit 14, for example, as described above, a bar candidate detection process, a bar identification process, a barcode decoding process, a decoding result determination process, and the like are performed. The bar recognition unit 14 outputs the processing result of the barcode recognition process to the outside through the output interface 16 (step S24).

  For example, when the barcode recognition is successful (when it is determined that the recognition result is valid information), the output interface 16 outputs the barcode recognition result to the outside. When the barcode recognition fails (when it is determined that the recognition result is not valid information), the output interface 16 outputs information indicating that the barcode recognition has failed to the outside. When the barcode recognition fails, as shown in a fifth processing example to be described later, the information obtained in each step of the barcode recognition processing is fed back to the image analysis unit 12 so that the image corresponding to the input image is displayed. The conversion process and the barcode recognition process may be executed again.

Next, a second processing example will be described.
In this second processing example, the image conversion method is determined based on the density value of the background (medium) in the 12-bit image. In the second to fifth processing examples described below, the image conversion unit 13 has a function of converting a 12-bit image into an 8-bit image based on a conversion table specified by the image analysis unit 12. It shall be.

  The medium density estimation unit 23 of the image analysis unit 12 has a function of determining a background image density value (medium density value) in a 12-bit image captured by the camera 2. The density value of the background image determined by the medium density estimation unit 23 is information indicating the characteristics of the background image in the input image. For example, the density value of the background image may be a mode density value or an average density value in the entire input image having 4096 gradations, or a density value determined by a percentile method.

FIG. 8 is a diagram illustrating an example of a density histogram for all pixels of an input image having 4096 gradations.
As shown in FIG. 8, the density histogram of an image obtained by photographing the barcode print surface on the medium by the camera 2 includes the density distribution of the background image and the density distribution of the bar image. In the present embodiment, as described above, it is assumed that each bar constituting the barcode is captured as an image brighter than the background. For this reason, the density distribution of the background image has a smaller density value than the density distribution of the bar image. That is, in the example of the histogram shown in FIG. 8, the distribution with the peak on the left side (the smaller density value) is the density distribution of the background image, and the distribution with the peak on the right side (the higher density value) is the bar image. The density distribution is determined. However, in an operation mode in which a dark barcode is printed on a light background (for example, an operation mode in which a black barcode is printed on a white background), the density distribution of the background image appears on the bright side, and the density distribution of the bar image Is estimated to appear on the dark side. For example, in the density value setting as shown in FIG. 8, it is estimated that the density distribution of the background image appears on the right side and the density distribution of the bar image appears on the left side, contrary to the density histogram shown in FIG.

Here, as shown in FIG. 8, it is assumed that the distribution having a peak on the left side in the density histogram is the density distribution of the background image, and the distribution having the peak on the right side is the density distribution of the bar image.
In the example of the histogram shown in FIG. 8, information indicating the characteristics of the distribution having a peak on the left side (that is, the density distribution of the background image) is determined as the density value (medium density value) of the background image. For example, the medium density estimation unit 23 determines a peak value (mode density value) or an average value of the distribution as information indicating the characteristics of the density distribution of the background image. The mode density value is determined by determining the peak value in the distribution on the left side of the histogram (the peak value is determined. The average value is the distribution of the background image determined from the peak value on the left side of the histogram. It is determined by calculating the average value of the density in the whole.

Further, the medium density estimation unit 23 may determine a density value at a position where the integrated value from the left side of the histogram exceeds a predetermined threshold as the medium density value determined by the percentile method. Here, an example in which the medium density value is determined by the percentile method as information indicating the characteristics of the density distribution of the background image will be described.
FIG. 9 is a diagram for explaining the medium density value determined by the percentile method.
When the medium density value is determined by the percentile method, the medium density estimation unit 23, as shown in FIG. 9, has a density distribution “0” at the left end in the background image density distribution appearance side as shown in FIG. ) To integrate. The medium density estimation unit 23 determines whether or not the calculated integral value is equal to or greater than a preset threshold value every time the integral value from the density value “0” to each density value is calculated. When the calculated integral value is equal to or greater than the threshold value, the medium density estimation unit 23 determines the density value that is the integral value as the medium density value.

  The threshold value for determining the medium density value by the above-mentioned percentile method should be set according to the assumed density distribution of the background image or the density distribution of the bar image. For example, in the example shown in FIG. 9, the density is set to be smaller than the density value that is the peak of the distribution determined as the density distribution of the background image. Further, the threshold value is set by, for example, a ratio (percentage) of an integrated value with respect to the entire density histogram.

  As described above, in the second processing example, the medium density estimation unit 23 determines a medium density value as information indicating the characteristics of the density distribution of the background image. When the medium density value is determined, the conversion method determination unit 26 determines an image conversion method according to the medium density value determined by the medium density estimation unit 23. In the second processing example, the conversion method determination unit 26 converts all pixels having a density value less than the medium density value to a density value “0”, and converts each pixel having a density value equal to or higher than the medium density value to 0 to 255. An image conversion method for linear conversion to the density value is determined. Such an image conversion method is designated to the image conversion unit 13 as a conversion table, for example.

FIG. 10 is a diagram for explaining an example of the conversion table Ta according to the medium density value.
As shown in FIG. 10, in the second processing example, the conversion method determination unit 26 creates a conversion table Ta that is a linear conversion based on the medium density value determined by the medium density estimation unit 23. In the image conversion process using the conversion table Ta as shown in FIG. 10, as described above, all pixels having a density value less than the medium density value in the input image are converted into the density value “0”, and the density value is changed in the input image. Each pixel having a medium density value or higher is linearly converted to a density value of 0 to 255.
According to the second processing example as described above, instead of dropping the information in the darker area than the medium density value, the information in the brighter area than the medium density value including the pixels constituting the bar image is displayed in a wide range of information ( The density value can be converted to 0-255).

Next, the flow of processing as the second processing example will be described.
FIG. 11 is a flowchart for explaining the flow of the second processing example in the barcode recognition apparatus 1.
When a 12-bit image captured by the camera 2 is acquired (step S31). The image interface 11 supplies the acquired 12-bit image to the image analysis unit 12 and the image conversion unit 13. When a 12-bit image is supplied from the image interface 11, the image analysis unit 12 performs a process of determining a medium density value in the 12-bit image by the medium density estimation unit 23 (steps S32 and S33).

  Here, the medium density value is determined by the percentile method as described above. That is, the medium density estimation unit 23 creates a density histogram indicating the density distribution of all pixels in the 12-bit image (step S32). When the histogram is created, the medium density estimation unit 23 determines a medium density value based on the histogram (step S33). As a process for determining the medium density value, the medium density estimation unit 23 calculates an integrated value from the density value “0” to each density value in order with respect to the histogram. Each time the integral value up to each density value is calculated, the medium density estimation unit 23 determines whether or not the integral value is equal to or greater than a predetermined threshold value. If it is determined that the integrated value is equal to or greater than a predetermined threshold, the medium density estimation unit 23 determines that the density value from which the integrated value is obtained is the medium density value in the 12-bit image.

  When the medium density value of the 12-bit image is determined by the medium density estimation unit 23, the conversion method determination unit 26 converts the 12-bit image into an 8-bit image for recognition processing based on the medium density value. The image conversion method is determined (steps S34 and S35). That is, the conversion method determination unit 26 creates a conversion table Ta for image conversion corresponding to the medium density value determined by the medium density estimation unit 23 (step S34). This conversion table Ta designates, for example, linear conversion that rises from the medium density value as a base point, as described above. When the conversion table Ta for image conversion is created, the conversion method determination unit 26 designates the created conversion table Ta as the image conversion method to the image conversion unit 13 (step S35).

When an image conversion method (conversion table) for the 12-bit image is specified by the conversion method determination unit 26 of the image analysis unit 12, the image conversion unit 13 converts the 12-bit image into an 8-bit image according to the specified conversion table Ta. Conversion processing is performed (step S36). The 8-bit image generated by the image conversion process based on the specified conversion table is supplied from the image conversion unit 13 to the bar recognition unit 14.
The bar recognition unit 14 performs barcode recognition processing on the 8-bit image supplied from the image conversion unit 13 (step S37). In the barcode recognition process in the bar recognition unit 14, for example, as described above, a bar candidate detection process, a bar identification process, a barcode decoding process, a decoding result determination process, and the like are performed. The bar recognition unit 14 outputs the processing result of the barcode recognition process to the outside through the output interface 16 (step S38).

  For example, when the barcode recognition is successful (when it is determined that the recognition result is valid information), the output interface 16 outputs the barcode recognition result to the outside. When the barcode recognition fails (when it is determined that the recognition result is not valid information), the output interface 16 outputs information indicating that the barcode recognition has failed to the outside. If the barcode recognition fails, an image corresponding to the input image is fed back to the image analysis unit 12 with information obtained in each step of the barcode recognition processing as in a fifth processing example described later. The conversion process and the barcode recognition process may be re-executed (retry).

Next, a third processing example will be described.
In the third processing example, the image conversion method is determined based on the density value of the bar constituting the barcode in the 12-bit image.
The bar density estimation unit 24 of the image analysis unit 12 has a function of determining a bar image density value (bar density value) in a 12-bit image (12-bit image) captured by the camera 2. The bar density value determined by the bar density estimation unit 24 is information indicating the characteristics of the bar image in the input image. For example, the bar density value may be a mode density value (peak value) or an average density value in the density distribution of the bar image, or may be a density value determined by a percentile method.

  In the present embodiment, as described above, it is assumed that each bar constituting the barcode is captured as an image brighter than the background. For this reason, the density distribution of the bar image has a density value larger than that of the background image. That is, in the example of the density histogram as shown in FIG. 8, it is estimated that the distribution having the peak on the right side (the one with the larger density value) is the density distribution of the bar image. However, in an operation mode in which a dark barcode is printed on a light background (for example, an operation mode in which a black barcode is printed on a white background), the density distribution of the bar image appears on the dark side, and the density of the background image The distribution appears on the bright side. For example, in the density value setting as shown in FIG. 8, the density distribution of the bar image appears on the left side and the density distribution of the background image appears on the right side, contrary to the histogram shown in FIG.

Here, as shown in FIG. 8, description will be made assuming that the distribution having a peak on the right side in the density histogram is the density distribution of the bar image.
In the example of the histogram shown in FIG. 8, information indicating the characteristics of the distribution having a peak on the right side (that is, the density distribution of the bar image) is determined as the density value (bar density value) of the bar image. For example, the bar density estimation unit 24 determines a mode density value or an average value of the distribution as information indicating the characteristics of the density distribution of the bar image. The mode density value is determined by determining a value (peak value) having the highest frequency in the distribution on the right side of the histogram. The average value is determined, for example, by calculating the average value of the density in the entire bar image distribution estimated from the peak value on the right side of the histogram.

The bar density estimation unit 24 determines the bar density value by the percentile method. The density value at the position where the integrated value from the right end of the histogram (density value “4096”) to each density value exceeds a predetermined threshold value. May be determined. Here, an example will be described in which the bar density value is determined by the percentile method as information indicating the characteristics of the density distribution of the bar image.
FIG. 12 is a diagram for explaining the bar density value determined by the percentile method.
When the bar density value is determined by the percentile method, the bar density estimation unit 24, as shown in FIG. 12, has a bar image density distribution side with respect to the histogram (in the example shown in FIG. Integration from density value “4096”) to each density value is performed. Each time the bar density estimation unit 24 calculates an integral value up to each density value, the bar density estimation unit 24 determines whether the calculated integral value is equal to or greater than a preset threshold value (threshold value for bar density value determination). . When the calculated integrated value is equal to or greater than the threshold value, the bar density estimating unit 24 determines the density value that has become the integrated value as the bar density value.

  The threshold for determining the bar density value by the percentile method should be set according to the assumed density distribution of the bar image or the density distribution of the background image. For example, in the example shown in FIG. 12, the bar density value is set to be a peak value in the density distribution of the bar image or a density value smaller than the minimum value of the density distribution of the bar image. Further, the threshold value is set by, for example, a ratio (percentage) of an integrated value with respect to the entire density histogram.

  As described above, in the third processing example, the bar density estimation unit 24 determines a bar density value as information indicating the characteristics of the density distribution of the bar image. When the bar density value is determined, the conversion method determination unit 26 determines an image conversion method according to the bar density value determined by the bar density estimation unit 24. In the third processing example, the conversion method determination unit 26 converts all pixels having a density value equal to or higher than the bar density value to a density value “255”, and converts each pixel having a density value less than the bar density value to 0 to 255. An image conversion method for linear conversion to the density value is determined. Information indicating such an image conversion method is designated to the image conversion unit 13 as a conversion table, for example.

FIG. 13 is a diagram for explaining an example of the conversion table Tb according to the bar density value.
As shown in FIG. 13, in the third processing example, the conversion method determination unit 26 creates a conversion table Tb that is a linear conversion based on the bar density value determined by the bar density estimation unit 24. In the image conversion process using the conversion table Tb as shown in FIG. 13, as described above, all pixels having a density value equal to or higher than the bar density value in the input image are converted to the density value “255”, and the density value is changed in the input image. Each pixel less than the bar density value is linearly converted to a density value of 0 to 255.
According to the third processing example as described above, instead of dropping the information of the brighter part than the bar density value (all are set to the maximum value), the bar density value including the density value of each pixel constituting the background image is used. The information of the dark part can be converted into information in a wide range (density value 0 to 255).

Next, a processing flow as the third processing example will be described.
FIG. 14 is a flowchart for explaining the flow of the third processing example in the barcode recognition apparatus 1.

  When a 12-bit image captured by the camera 2 is acquired (step S41). The image interface 11 supplies the acquired 12-bit image to the image analysis unit 12 and the image conversion unit 13. When a 12-bit image is supplied from the image interface 11, the image analysis unit 12 performs a process of determining a bar density value in the 12-bit image by the medium density estimation unit 23 (steps S42 and S43). Here, it is assumed that the bar density value is determined by the percentile method as described above. That is, the bar density estimation unit 24 creates a histogram indicating the density distribution of all pixels in the 12-bit image (step S42). When the histogram is created, the bar density estimation unit 24 determines the bar density value based on the histogram (step S43). That is, the bar density estimation unit 24 calculates an integrated value from the maximum density value “4095” to each density value in order with respect to the histogram. Each time the integrated value up to each density value is calculated, the bar density estimating unit 24 determines whether or not the integrated value is equal to or greater than a predetermined threshold value. If it is determined that the integrated value is greater than or equal to a predetermined threshold value, the bar density estimation unit 24 determines that the density value from which the integrated value is obtained is the bar density value in the 12-bit image.

  When the bar density value of the 12-bit image is determined by the bar density estimating unit 24, the conversion method determining unit 26 converts the 12-bit image into an 8-bit image for recognition processing based on the bar density value. The image conversion method is determined (steps S44 and S45). That is, the conversion method determination unit 26 creates a conversion table Tb for image conversion corresponding to the bar density value determined by the bar density estimation unit 24 (step S44). For example, as described above, this conversion table specifies linear conversion from the minimum density value (“0”) to the bar density value. When the conversion table Tb for image conversion is created, the conversion method determination unit 26 designates the created conversion table Tb as the image conversion method to the image conversion unit 13 (step S45).

When an image conversion method (conversion table) for the 12-bit image is specified by the conversion method determination unit 26 of the image analysis unit 12, the image conversion unit 13 converts the 12-bit image into an 8-bit image according to the specified conversion table Tb. Conversion processing is performed (step S46). The 8-bit image generated by the image conversion process based on the specified conversion table is supplied from the image conversion unit 13 to the bar recognition unit 14.
The bar recognition unit 14 performs barcode recognition processing on the 8-bit image supplied from the image conversion unit 13 (step S47). In the barcode recognition process in the bar recognition unit 14, for example, as described above, a bar candidate detection process, a bar identification process, a barcode decoding process, a decoding result determination process, and the like are performed. The bar recognition unit 14 outputs the processing result of the barcode recognition process to the outside through the output interface 16 (step S48).

  For example, when the barcode recognition is successful (when it is determined that the recognition result is valid information), the output interface 16 outputs the barcode recognition result to the outside. When the barcode recognition fails (when it is determined that the recognition result is not valid information), the output interface 16 outputs information indicating that the barcode recognition has failed to the outside. If the barcode recognition fails, an image corresponding to the input image is fed back to the image analysis unit 12 with information obtained in each step of the barcode recognition processing as in a fifth processing example described later. The conversion process and the barcode recognition process may be re-executed (retry).

Next, a fourth processing example will be described.
In the fourth processing example, an image conversion method is determined based on a combination of a background (medium) density value and a bar density value in an input image. In other words, the fourth processing example creates a conversion table for designating an image conversion method by combining the medium density value obtained in the second processing example and the bar density value obtained in the third processing example. Is.

  That is, in the fourth processing example, the medium density estimation unit 23 determines the medium density value as described in the second processing example with respect to the 12-bit image captured by the camera 2, and further, the bar The density estimation unit 24 determines the bar density value as described in the third processing example.

  For example, it is assumed that the medium density value as shown in FIG. 9 and the bar density value as shown in FIG. 12 are determined from the density histogram as shown in FIG. In this case, the conversion table Ta based on the medium density value is obtained by linearly converting the medium density value to the maximum density value as shown in FIG. As shown in FIG. 13, the conversion table Tb based on the bar density value is obtained by linear conversion from the minimum density value to the bar density value. Therefore, in the fourth processing example, it is considered that a conversion table Tab considering the medium density value and the bar density value can be obtained by combining the conversion table Ta and the conversion table Tb.

  That is, the conversion method determining unit 26 determines the conversion table Tab based on the medium density value determined by the medium density estimating unit 23 and the bar density value determined by the bar density estimating unit 24.

  Various methods are conceivable for determining the conversion table Tab based on the combination of the medium density value and the bar density value. The conversion table Tab only needs to emphasize density changes from the medium density value to the bar density value. For example, the conversion table Tab may be created by adding and averaging the conversion table Ta based on the medium density value and the conversion table Tb based on the bar density value, or from the medium density value to the bar density value. A conversion table that linearly converts the density values up to the above may be used. In the former case, image conversion processing that reflects density changes at density values less than the medium density value and density values above the bar density value while emphasizing density changes from the media density value to the bar density value is performed. it can. In the latter case, it is possible to perform image conversion in which the density change from the medium density value to the bar density value is enhanced to the maximum (for example, converted from “0” to “255” in 256 gradations). Furthermore, the conversion table Tab may be determined by combining the former and the latter.

Next, a processing flow as the fourth processing example will be described.
FIG. 15 is a flowchart for explaining the flow of the fourth processing example in the barcode recognition apparatus 1.

  When a 12-bit image captured by the camera 2 is acquired (step S51). The image interface 11 supplies the acquired 12-bit image to the image analysis unit 12 and the image conversion unit 13. When a 12-bit image is supplied from the image interface 11, the image analysis unit 12 determines the medium density value in the 12-bit image by the medium density estimation unit 23 (steps S52 and S53), and the bar density estimation unit. 24 determines the bar density value in the 12-bit image (step S54).

  When the medium density value and the bar density value of the 12-bit image are determined, the conversion method determining unit 26 converts the 12-bit image into an 8-bit image for recognition processing based on the medium density value and the bar density value. An image conversion method for determining the image is determined (steps S55 and S56). That is, the conversion method determination unit 26 creates a conversion table Tab for image conversion according to the medium density value and the bar density value (step S55). For example, as described above, the conversion table Tab designates conversion that emphasizes the density change from the medium density value to the bar density value. When the conversion table Tab for image conversion is created, the conversion method determination unit 26 designates the created conversion table Tab as the image conversion method to the image conversion unit 13 (step S56).

When an image conversion method (conversion table) for the 12-bit image is specified by the conversion method determination unit 26 of the image analysis unit 12, the image conversion unit 13 converts the 12-bit image into an 8-bit image according to the specified conversion table Tab. Conversion processing is performed (step S57). The 8-bit image generated by the image conversion process based on the specified conversion table is supplied from the image conversion unit 13 to the bar recognition unit 14.
The bar recognition unit 14 performs barcode recognition processing on the 8-bit image supplied from the image conversion unit 13 (step S58). In the barcode recognition process in the bar recognition unit 14, for example, as described above, a bar candidate detection process, a bar identification process, a barcode decoding process, a decoding result determination process, and the like are performed. The bar recognition unit 14 outputs the processing result of the barcode recognition process to the outside through the output interface 16 (step S59).

  For example, when the barcode recognition is successful (when it is determined that the recognition result is valid information), the output interface 16 outputs the barcode recognition result to the outside. When the barcode recognition fails (when it is determined that the recognition result is not valid information), the output interface 16 outputs information indicating that the barcode recognition has failed to the outside. If the barcode recognition fails, an image corresponding to the input image is fed back to the image analysis unit 12 with information obtained in each step of the barcode recognition processing as in a fifth processing example described later. The conversion process and the barcode recognition process may be re-executed.

Next, a fifth processing example will be described.
In the fifth processing example, the image conversion process and the barcode recognition process are re-executed (retry) based on the information fed back to the image analysis unit 12 by the feedback unit 15. In the fifth processing example, the image analysis unit 12 has a function of determining an image conversion method based on the fed back information, and the image conversion unit 13 converts the input image into an image for recognition processing again. It has a re-conversion function, and the bar recognition unit 14 has a retry function for executing the barcode recognition process again based on the image for recognition processing obtained by the re-conversion function.

  That is, in the fifth processing example, when it is determined that the barcode recognition result by the bar recognition unit 14 is not valid, or when the barcode recognition result by the bar recognition unit 14 is determined to be erroneous, The feedback unit 15 feeds back the information obtained in the bar recognition processing step or the information generated from the information obtained in the bar recognition processing step to the image analysis unit 12 as feedback information. The image analysis unit 12 re-determines an image conversion method based on the fed back information. In this case, the image conversion unit 13 executes a process of converting the input image into an image for recognition processing again according to an image conversion method based on the feedback information specified by the image analysis unit 12. The bar recognition unit 14 performs barcode recognition processing again on the reconverted image for recognition processing.

  Next, the feedback information will be described.

  As described above, the feedback information is information for determining an image conversion method for reconverting an input image into an image for recognition processing. Therefore, the feedback information needs to be information for determining an image conversion method that can accurately recognize a barcode. Here, it is assumed that the feedback information is information obtained in the bar recognition processing step or information determined from the information obtained in the bar recognition processing step. When feedback information is determined from the information obtained in the bar recognition process, the feedback information determination process may be executed by the bar recognition unit 14 or the feedback unit 15. Alternatively, the image analysis unit 12 may execute the process.

Next, a first example of feedback information will be described.
First, as a first example of feedback information, a process (feedback information determination process) for determining an optimum density value (separation threshold) for separating a background image and a bar image will be described. The separation threshold is information determined from information obtained in the bar recognition process. Here, it is assumed that the feedback unit 15 determines the separation threshold from the information obtained in the bar recognition process.

  The bar recognition unit 14 performs bar candidate detection processing as described above as one step of barcode recognition processing. In the bar candidate detection process, when the density value of the background image and the density value of the bar image are close to each other in the image for recognition processing, it is difficult to reliably detect all the bars as bar candidates. In other words, if the input image is converted into an image for recognition processing in which the density difference between the density value of the background image and the density value of the bar image is emphasized, the bar candidate detection accuracy can be improved. The separation threshold as the feedback information is information for determining an image conversion method in which the density difference between the background image and the bar image is emphasized.

  In many cases, the bar recognizing unit 14 can detect some bars even if all bars cannot be detected in the bar candidate detection process. In such a case, the bar recognition unit 14 supplies information indicating the detected bar candidate area to the feedback unit 15. The feedback unit 15 determines the separation threshold based on the image of the detected bar candidate region in the input image.

  For example, the feedback unit 15 determines the bar density value and the density value of the background image from the image of the bar candidate area, and determines the intermediate value as the separation threshold. In this case, the feedback unit 15 extracts an image of a bar candidate area from the input image, and uses a density value such as a mode density value or an average density value obtained from the extracted bar candidate area image as a bar candidate image. The density value is determined. As the density value of the background image, as described in the first to fourth processing examples, it is possible to use the mode density value obtained from the entire input image, the average value, or the density value determined by the percentile method. It is.

  Note that when the shape of the bars and the arrangement of the bars (for example, the interval between the bars) are in a predetermined form, the background image area may be estimated from the detected bar candidate image areas. In this case, the feedback unit 15 estimates the background image area from the bar candidate image area, extracts the background image area image estimated from the input image, and obtains the mode density value obtained from the extracted background image. Alternatively, a density value such as an average density value may be determined as the density value of the background image.

  As described above, when at least one bar candidate is detected, it is possible to estimate the bar density value or the background image density value in the input image based on information indicating the bar candidate region. . When at least one bar is detected as a bar candidate, it is considered that the bar density value obtained from the bar candidate area is more accurate than the bar density value determined from the density histogram of the entire image.

  The separation threshold determined based on the information indicating the bar candidate area as described above is fed back to the image analysis unit 12. When a separation threshold value is supplied as feedback information, the image analysis unit 12 creates a conversion table based on the fed back separation threshold value. The conversion table based on the separation threshold specifies an image conversion method for obtaining an image for recognition processing in which the density difference between the background image and the bar image is emphasized. That is, the image analysis unit 12 creates a conversion table that converts density values near the separation threshold in the input image into a wide range of density values.

  In the image conversion process based on such a conversion table, an image for recognition processing (for example, 8-bit image) in which the density change near the separation threshold in the input image (for example, 12-bit image) is emphasized is obtained. That is, the separation threshold is a density value that separates the background image and the bar image. For this reason, it is considered that the image conversion process based on the conversion table provides a recognition process image in which the density difference between the background image and the bar image is emphasized.

FIG. 16 is a diagram for explaining an example of the conversion table based on the separation threshold.
In the example shown in FIG. 16, the density value in a predetermined range before and after the separation threshold is converted into a wide range of density values, and the other density values are converted into density values in a narrow range. For example, in the conversion table as shown in FIG. 16, when an input image with 4096 gradations is converted into an image for recognition processing with 256 gradations, pixels whose density values are separation threshold −100 to separation threshold +100 are density values 27. To a pixel having a density value of −100 or less, and a pixel having a density value of the separation value + 100 or more is converted to a pixel having a density value of 227 to 255. You can specify image conversion to convert to. In this case, an image for recognition processing in which the density change before and after the separation threshold in the input image is emphasized is obtained.

Next, a second example of feedback information will be described.
As a second example of feedback information, a process for determining a density histogram in a barcode area candidate will be described. The barcode area candidate is information obtained in the bar recognition process. It is expected that the background area noise is less in the barcode area candidate image in the input image than in the entire input image. Background noise is, for example, an image irrelevant to a background image or a barcode, and for example, a printed advertisement on a printed surface of a percode is assumed. That is, the density histogram for the barcode area candidate image in the input image is more likely to reduce the influence of background noise than the density histogram for the entire input image.

  The bar recognition unit 14 performs a process of detecting a barcode area candidate from the detected bar candidates as one step of the barcode recognition process. For example, the bar recognition unit 14 can detect a bar code area or a part of the bar code area as long as a part of the bars can be detected even if all the bars cannot be detected. Many. In such a case, the bar recognition unit 14 supplies information indicating the detected barcode area to the image analysis unit 12 via the feedback unit 15. Thereby, the image analysis unit 12 can generate a density histogram only for the image around the barcode, and can determine an image conversion method according to the density histogram.

  As described in the first to fourth processing examples, the image analysis unit 12 has a function of creating a density histogram and determining a mode density value, a medium density value, a bar density value, or the like from the created density histogram. Have. When the information indicating the barcode area is obtained from the bar recognition unit 14, the image analysis unit 12 targets the image of the barcode area in the input image as described in the first to fourth processing examples. Processing is possible. That is, the image analysis unit 12 can determine the image conversion method based on the mode density value, medium density value, bar density value, or combination of medium density value and bar density value obtained from the barcode area. .

Next, the flow of processing as the fifth processing example will be described.
FIG. 17 is a flowchart for explaining the flow of the fifth processing example in the barcode recognition apparatus 1.
When an input image (for example, a 4096 gradation 12-bit image) captured by the camera 2 is acquired as an input image (step S61). The image interface 11 supplies the acquired input image to the image analysis unit 12 and the image conversion unit 13. When an input image is supplied from the image interface 11, the image analysis unit 12 performs an image analysis process for determining features obtained from the input image (step S62). In the image analysis processing in step S62, the image analysis unit 12 determines the mode density value, the medium density value, or the bar density value from the entire input image as information for determining the image conversion method. An image conversion method such as a conversion table is determined based on the determined information. For example, the processing described in the second to fourth processing examples (steps S11 to S21, S32 to S35, S42 to S45, and S52 to S56) can be applied as the processing of step S62. For this reason, detailed description of the processing in step S62 is omitted.

  When the image conversion method is specified by the image analysis unit 12, the image conversion unit 13 converts the input image into an image for recognition processing (for example, 8 bits of 256 gradations) according to the image conversion method specified by the image analysis unit 12. (Image) is converted (step S63). An image for recognition processing obtained as a result of this image conversion processing is supplied from the image conversion unit 13 to the bar recognition unit 14. The bar recognition unit 14 performs barcode recognition processing on the image for recognition processing supplied from the image conversion unit 13 (step S64).

  Note that the image conversion process of step S63 executed subsequent to steps S61 and S62 corresponds to the image conversion process by the first image conversion method, and the image conversion unit 13 that executes this is the first image conversion unit. It corresponds. Further, the barcode recognition process of step S64 executed subsequent to steps S61 to S63 corresponds to the first barcode recognition process, and the bar recognition unit 14 that executes this corresponds to the first bar recognition unit. ing.

  In the barcode recognition process in the bar recognition unit 14, for example, as described above, a bar candidate detection process, a bar identification process, a barcode decoding process, a decoding result determination process, and the like are performed. Here, assuming that the process up to obtaining the result of the barcode decoding process is the barcode recognition process, the bar recognition unit 14 has valid information obtained by the barcode recognition process (decoding result). Whether or not the recognition result (decoding result) is determined is determined (step S65).

  In the recognition result determination process, it is determined whether the recognition result of the barcode recognition process is desired information. For example, whether or not the recognition result is valid is determined based on whether or not the barcode recognition result is information of a predetermined number of digits. Further, whether or not the recognition result is valid may be determined based on whether or not the barcode recognition result matches information registered in a dictionary database (not shown). When it is determined that the recognition result obtained by the barcode recognition process is valid (YES in step S65), the bar recognition unit 14 outputs the processing result of the barcode recognition process to the outside through the output interface 16. (Step S66).

  When it is determined that the recognition result obtained by the barcode recognition process is not valid (NO in step S65), the bar recognition unit 14 further determines whether or not the barcode recognition process for the image can be retried. (Step S67). For the retry of the barcode recognition process, for example, the number of retries may be limited, or the total processing time for the input image may be limited. If it is determined that the retry is impossible by the above determination (NO in step S67), the bar recognition unit 14 outputs information indicating that the barcode recognition has failed through the output interface 16 to the outside.

  If it is determined that retry is possible based on the above determination (step S67, YES), the bar recognition unit 14 determines to re-execute the barcode recognition processing for the image. When it is determined that the barcode recognition process is to be re-executed, the bar recognition unit 14 or the feedback unit 15 uses the information obtained in the process of the bar recognition process for the image or information determined from the information as feedback information. A process of feeding back to the image analysis unit 12 is performed (steps S68 and S69). Here, it is assumed that the feedback unit 15 determines feedback information from information obtained in the process of bar recognition processing. In this case, the feedback unit 15 acquires information obtained during the bar recognition process from the bar recognition unit 14, and determines feedback information (step S68). For example, when the separation threshold between the background and the bar is used as feedback information, the feedback unit 15 acquires information indicating the bar candidate area from the bar recognition unit 14 as described above, and the acquired bar candidate area The separation threshold is determined based on the indicated information.

  When the feedback information is determined, the feedback unit 15 feeds back the feedback information to the image analysis unit 12. When the feedback information is fed back, the image analysis unit 12 performs an image analysis process on the input image based on the feedback information (step S70). In this image analysis process, an image conversion method for the input image is determined based on the feedback information. The image conversion method determined based on such feedback information is designated to the image conversion unit 13. When the image conversion method based on the feedback information is designated to the image conversion unit 13, the image conversion unit 13, the bar recognition unit 14, and the like again execute the processing from step S63 described above.

  Note that the image conversion process of step S63 executed subsequent to steps S68-S70 corresponds to the image conversion process by the second image conversion method, and the image conversion unit 13 that executes this is the second image conversion unit. It corresponds. In addition, the barcode recognition process in step S64 executed subsequent to steps S68-S70 and S63 corresponds to the second barcode recognition process, and the bar recognition unit 14 that executes the barcode recognition process becomes the second bar recognition unit. It corresponds.

  In the fifth processing example as described above, a valid barcode recognition result is not obtained in the predetermined barcode recognition processing, but based on information obtained in the process of barcode recognition processing. By analyzing the input image, an optimal image conversion method for converting the input image into an image for recognition processing can be determined. That is, in the fifth processing example, even if a valid recognition result cannot be obtained by the first-stage barcode recognition process, the second stage can obtain a barcode recognition result with high accuracy. It is possible to execute the barcode recognition process.

  In the fifth processing example, a high-precision barcode recognition process is executed by the second-stage barcode recognition process on the input image that cannot be recognized by the first-stage barcode recognition process. It has become. Therefore, in the fifth processing example, the first stage barcode recognition process is not limited to the above-described processing procedure, and can be performed by various processing procedures. For example, as a fifth processing example, step S62 may be omitted, and the first-stage barcode recognition process may be performed using an image converted by a predetermined image conversion method. Also in this case, in the second stage barcode recognition process, the information obtained in the first stage barcode recognition process is applied to the input image that has failed to be recognized in the first stage barcode recognition process. The input image can be converted into an image for recognition processing by the image conversion method optimized by using the method, and the barcode recognition processing can be executed on the converted image.

[1]
An image acquisition unit for acquiring an input image including an image to be detected;
An image analysis unit that analyzes an input image acquired by the image acquisition unit and determines an image conversion method for converting the input image into an image in a format for recognition processing based on the analysis result;
An image conversion unit that converts an input image acquired by the image acquisition unit into an image for recognition processing by the image conversion method determined by the image analysis unit;
A detection target recognition unit that performs recognition processing of a detection target on the image for recognition processing obtained by the image conversion unit;
An object recognition apparatus comprising:

[2]
The image conversion unit has a function of converting an input image into an image for recognition processing by a plurality of types of image conversion methods;
The image analysis unit selects at least one image conversion method from a plurality of types of image conversion methods that the image conversion unit has, based on the analysis result of the input image acquired by the image acquisition unit.
The object recognition device according to [1] above, wherein

[3]
The image analysis unit determines a density value of a background image included in the input image acquired by the image acquisition unit, and determines an image conversion method for the input image based on the determined density value of the background image;
The object recognition device according to [1] above, wherein

[4]
The image analysis unit determines a density value of a detection element image constituting a detection target included in the input image acquired by the image acquisition unit, and performs image conversion on the input image based on the determined density value of the detection element image Decide how,
The object recognition device according to [1] above, wherein

[5]
The image analysis unit determines a density value of a background image included in the input image acquired by the image acquisition unit and a density value of a detection element image constituting a detection target, and determines the density value of the determined background image and the detection element Determining an image conversion method for the input image based on the density value of the image;
The object recognition device according to [1] above, wherein

[6]
An image acquisition unit for acquiring an input image including an image to be detected;
A first image conversion unit that converts the input image acquired by the image acquisition unit into an image in a format for recognition processing;
A first detection target recognition unit that performs a detection target recognition process on the recognition processing image obtained by the first image conversion unit;
When recognition of the detection target by the first detection target recognition unit fails, the input image is analyzed based on information obtained in the process of recognition processing of the detection target by the first detection target recognition unit, An image analysis unit that determines an image conversion method for reconverting the input image into an image for recognition processing based on the analysis result;
A second image conversion unit that reconverts the input image into an image for recognition processing based on an analysis result by the image analysis unit;
A second detection target recognition unit that performs a detection target recognition process on the recognition processing image obtained by the second image conversion unit;
An object recognition apparatus comprising:

[7]
A target recognition method used in a target recognition device,
Obtain an input image containing the image to be detected,
Image analysis of the acquired input image,
Determining an image conversion method for converting the input image into an image of a format for recognition processing based on an analysis result of the input image;
Converting the input image into an image for recognition processing by the determined image conversion method;
Recognizing the detection target for the image for recognition processing obtained by the conversion,
An object recognition method characterized by the above.

[8]
As the image conversion, a plurality of types of image conversion methods for converting an input image into an image for recognition processing can be executed.
The determination of the image conversion method is to select at least one image conversion method from a plurality of types of image conversion methods that can be executed as the image conversion based on the analysis result of the input image.
The object recognition method according to [7], wherein the object is recognized.

[9]
The image analysis determines a density value of a background image included in the input image,
The determination of the image conversion method determines an image conversion method for the input image based on a density value of a background image determined by the image analysis.
The object recognition method according to [7], wherein the object is recognized.

[10]
The image analysis determines a density value of a detection element image constituting a detection target included in the input image,
The determination of the image conversion method determines an image conversion method for the input image based on a density value of the detection element image determined by the image analysis.
The object recognition method according to [7], wherein the object is recognized.

[11]
The image analysis determines a density value of a background image included in the input image and a density value of a detection element image constituting a detection target;
The determination of the image conversion method determines an image conversion method for the input image based on the density value of the background image and the density value of the detection element image determined by the image analysis.
The object recognition method according to [7], wherein the object is recognized.

[12]
A target recognition method used in a target recognition device,
Obtain an input image containing the image to be detected,
Converting the input image into an image in a format for recognition processing by the first image conversion method;
A recognition target recognition process is performed on the recognition process image obtained by the first image conversion method,
When the recognition of the detection target fails, the input image is analyzed based on the information obtained in the process of recognition processing of the detection target,
Determining a second image conversion method for reconverting the input image into an image for recognition processing based on the analysis result of the input image;
Re-converting the input image into an image for recognition processing by the second image conversion method;
Re-execute the recognition processing of the detection target on the recognition processing image obtained by the reconversion,
An object recognition method characterized by the above.

  DESCRIPTION OF SYMBOLS 1 ... Barcode recognition apparatus, 2 ... Camera, 11 ... Image interface, 12 ... Image analysis part, 13 ... Image conversion part, 14 ... Bar recognition part, 15 ... Feedback part, 16 ... Output interface, 21 ... Density value estimation part , 22 ... Conversion method determining unit, 23 ... Medium density estimating unit, 24 ... Bar density estimating unit, 26 ... Conversion method determining unit, 31 ... Control device, 32 ... Feeding device, 33 ... Conveying device, 34 ... Sorting device, 41 ... Bar density estimation part, 42 ... Medium density estimation part

Claims (15)

An image acquisition unit for acquiring an input image including an image to be detected;
A first detection target recognition unit that performs detection target recognition processing on the input image acquired by the image acquisition unit;
When recognition of the detection target by the first detection target recognition unit fails, the input image is used for recognition processing based on information obtained in the process of recognition processing of the detection target by the first detection target recognition unit. An image analysis unit for determining an image conversion method for re-converting into an image;
An image conversion unit that converts the input image into an image for recognition processing by the image conversion method determined by the image analysis unit;
A second detection target recognition unit that performs a detection target recognition process on the recognition processing image obtained by the image conversion unit;
A target recognition device having A feedback unit that determines feedback information from information obtained in the process of recognition processing of the detection target in the first detection target recognition unit when recognition of the detection target by the first detection target recognition unit fails; ,
The image analysis unit determines an image conversion method based on feedback information determined by the feedback unit;
The subject recognition apparatus according to claim 1. The feedback information is a separation threshold for separating a background image in the input image and a detection element image constituting a detection target.
The subject recognition apparatus according to claim 2. The image analysis unit determines an image conversion method based on the separation threshold as the feedback information;
The subject recognition apparatus according to claim 3. The image analysis unit determines an image conversion method based on a conversion table created based on the separation threshold as the feedback information;
The subject recognition apparatus according to claim 4. The feedback information is a density histogram in an image of an image region candidate to be detected in the input image.
The subject recognition apparatus according to claim 2. The image analysis unit determines an image conversion method for reconverting the input image into an image for recognition processing based on a plurality of types of density values detected from the density histogram as the feedback information.
The object recognition apparatus according to claim 6, wherein: A target recognition method used in a target recognition device,
Obtain an input image containing the image to be detected,
Performing a first recognition process for recognizing a detection target for the input image;
An image conversion method for converting the input image into an image for recognition processing based on information obtained in the course of the first recognition processing when recognition of a detection target by the first recognition processing fails. Decide
The input image is converted into an image for recognition processing by the determined image conversion method,
Performing a second recognition process for recognizing a detection target on an image for recognition processing obtained by image conversion by the image conversion method;
Object recognition method. Further, when the recognition of the detection target by the first recognition process fails, feedback information for determining the image conversion method is determined from information obtained in the course of the first recognition process,
The image conversion method is determined based on the determined feedback information.
The object recognition method according to claim 1. The feedback information is a separation threshold for separating a background image in the input image and a detection element image constituting a detection target.
The object recognition method according to claim 2. The image conversion method is determined based on the separation threshold as the feedback information.
The object recognition method according to claim 3. The image conversion method is determined based on a conversion table created based on the separation threshold as the feedback information.
The subject recognition method according to claim 4. The feedback information is a density histogram in an image of an image region candidate to be detected in the input image.
The object recognition method according to claim 2. The image conversion method is determined based on a plurality of types of density values detected from the density histogram as the feedback information.
The object recognition method according to claim 6. An image analysis unit that determines an image conversion method based on information obtained by recognition processing of a detection target for an input image;
A detection target recognition unit that performs recognition processing of a detection target on an input image converted by the image conversion method determined by the image analysis unit;
A target recognition device having

Images